Solution trivially zero and failed to allocate memory

Numerical methods and mathematical models of Elmer
Post Reply
guillem_arque
Posts: 13
Joined: 09 Mar 2021, 22:00
Antispam: Yes

Solution trivially zero and failed to allocate memory

Post by guillem_arque »

Hello everyone,
I am having some trouble solving a magnetodynamics problem.
The thing is that it appears the message of solution trivially zero, even though previous simulations worked well because I had done the modifications said in another post about this topic. I post my case file in case anyone can see the error and help me.
Moreover, another difference with previous cases that I have simulated, the mesh is bigger and in the solver log appears the message of GCR: failed to allocate memory of size: 1868765 x 500.
I really don't know what to do to solve this problems.
Thank you
Guillem
Attachments
solverlog.txt
(4.03 KiB) Downloaded 10 times
case.sif
(3.66 KiB) Downloaded 12 times
raback
Site Admin
Posts: 4066
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Solution trivially zero and failed to allocate memory

Post by raback »

Hi

I don't see any boundary conditions for the vector potential as in
https://github.com/ElmerCSC/elmerfem/bl ... e/case.sif

GCR allocates vectors for all 500 iterations. Rather try BiCGStabl or Idrs and then you can use much more iterations too. The BCs may help in convergence.

-Peter
guillem_arque
Posts: 13
Joined: 09 Mar 2021, 22:00
Antispam: Yes

Re: Solution trivially zero and failed to allocate memory

Post by guillem_arque »

Hi Peter,
I really appreciate your response.
I have added BCs to my case and the objective is to introduce current into the wire (the case is a cube of air and thisconductor in the middle), but I still get solution trivially zero, with BiCGStabl.
If yo know what my error is it woul be really helpful.
Thank you and sorry for the inconveniences.
Guillem
Attachments
case.sif
(3.66 KiB) Downloaded 8 times
kevinarden
Posts: 1027
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Solution trivially zero and failed to allocate memory

Post by kevinarden »

I believe he is saying that the Electric Infinity BC = True is defined in the StatElecSolver
but not in the WhitneyAVSolver, the outer BC is the WhitneyAVSolver is like:

Boundary Condition 6
Name = "AirEnd"
Target Boundaries(1) = 6

AV re {e} 1 = Real 0.0
AV re {e} 2 = Real 0.0

AV im {e} 1 = Real 0.0
AV im {e} 2 = Real 0.0
End
guillem_arque
Posts: 13
Joined: 09 Mar 2021, 22:00
Antispam: Yes

Re: Solution trivially zero and failed to allocate memory

Post by guillem_arque »

And how do you specify to the Outer BC that it is an Infinite BC, like in the Electrostatic Solver?
kevinarden
Posts: 1027
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Solution trivially zero and failed to allocate memory

Post by kevinarden »

My understanding is that is what AV = 0 is.
guillem_arque
Posts: 13
Joined: 09 Mar 2021, 22:00
Antispam: Yes

Re: Solution trivially zero and failed to allocate memory

Post by guillem_arque »

Perfect, thank you!
Now it works fine , but only introducing current density in real numbers.
Does anyone know how to introduce imaginary components of current density in the WhitneyAVSolver?
Guillem
raback
Site Admin
Posts: 4066
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Solution trivially zero and failed to allocate memory

Post by raback »

Hi

The "infinite bc" is a Robin type of boundary condition that may be derived when you assume that your field goes down as in spherical symmetry. It is a useful approximation for diffusion type of equations far from the source.

Now for AV solver you cannot use this approximation. This is because 1) the \curl\curl operator behaves differently that the \div\grad operator (laplace) and 2) for the edge element formulation you don't even have any dof normal to the surface that we could set.

Hence, what Kevin says is the best you can do.

-Peter
guillem_arque
Posts: 13
Joined: 09 Mar 2021, 22:00
Antispam: Yes

Re: Solution trivially zero and failed to allocate memory

Post by guillem_arque »

Perfect!
It works perfectly now, thank you so much.
Now I just need to discover if it is possible to introduce current with real and imaginary components.
I really appreciate your answers!
Guillem
Post Reply